The cost of building an AI product like ChatGPT can be difficult to measure. But one thing it needs for its technology is OpenAI it is plenty of water, which was drawn from the Raccoon and Des Moines River watersheds in central Iowa.
Water cools a powerful supercomputer as it tries to teach AI systems how to mimic human handwriting.
As they scramble to capitalize on the artificial intelligence frenzy, leading tech developers from Microsoft, OpenAI and Google have acknowledged that growing demand for AI tools comes with heavy costs, from expensive semiconductors to excessive water consumption.
Often, of course, companies are very secretive about these specifics. Few people in Iowa knew the state was the birthplace of OpenAI's most advanced model, GPT-4, before a top Microsoft executive said in a speech that it was "literally made next to a cornfield west of Des Moines."
Building a large language model requires analyzing patterns from a huge repository of human-written texts.
All this calculation requires a lot of electricity and of course produces a lot of heat. To keep cool on hot days, data centers must constantly pump water.
At its latest environmental report, Microsoft revealed that global water consumption rose 34% from 2021 to 2022 (to nearly 1,7 billion gallons, or more than 2.500 Olympic-sized swimming pools), a sharp increase compared to previous years that many outside researchers associate with the company's AI research.
"We should say the increase is due to artificial intelligence," including the company's "big investment in genetic artificial intelligence and its collaboration with OpenAI," said Shaolei Ren, a researcher at the University of California, Riverside who was trying to calculate the environmental impact of productive artificial intelligence products such as ChatGPT;
In a paper to be published later this year, Ren's team estimates that ChatGPT swallows 500 milliliters of water every time you ask it a series of 5 to 50 questions. .
Google Reported a 20% increase in water use over the same period, which Ren also attributes largely to the company's AI work.
OpenAI and Microsoft have both said they are working to improve the "efficiency" of training AI models.
George is still wondering what he is doing here….
